A Bayesian comparative effectiveness trial in action: developing a platform for multisite study adaptive randomization

In the last few decades, the number of trials using Bayesian methods has grown rapidly. Publications prior to 1990 included only three clinical trials that used Bayesian methods, but that number quickly jumped to 19 in the 1990s and to 99 from 2000 to 2012. While this literature provides many examples of Bayesian Adaptive Designs (BAD), none of the papers that are available walks the reader through the detailed process of conducting a BAD. This paper fills that gap by describing the BAD process used for one comparative effectiveness trial (Patient Assisted Intervention for Neuropathy: Comparison of Treatment in Real Life Situations) that can be generalized for use by others. A BAD was chosen with efficiency in mind. Response-adaptive randomization allows the potential for substantially smaller sample sizes, and can provide faster conclusions about which treatment or treatments are most effective. An Internet-based electronic data capture tool, which features a randomization module, facilitated data capture across study sites and an in-house computation software program was developed to implement the response-adaptive randomization.

A process for adapting randomization with minimal interruption to study sites was developed. A new randomization table can be generated quickly and can be seamlessly integrated in the data capture tool with minimal interruption to study sites.

This manuscript is the first to detail the technical process used to evaluate a multisite comparative effectiveness trial using adaptive randomization. An important opportunity for the application of Bayesian trials is in comparative effectiveness trials. The specific case study presented in this paper can be used as a model for conducting future clinical trials using a combination of statistical software and a web-based application.

ClinicalTrials.gov Identifier: NCT02260388, registered on 6 October 2014